Uncertainty Estimation and Analysis of Categorical Web Data

Web data often manifest high levels of uncertainty. We focus on categorical Web data and we represent these uncertainty levels as firstor second-order uncertainty. By means of concrete examples, we show how to quantify and handle these uncertainties using the BetaBinomial and the Dirichlet-Multinomial models, as well as how take into account possibly unseen categories in our samples by using the Dirichlet process. We conclude by exemplifying how these higher-order models can be used as a basis for analyzing datasets, once at least part of their uncertainty has been taken into account. We demonstrate how to use the Battacharyya stastistical distance to quantify the similarity between Dirichlet distributions, and use such results to analyze a Web dataset of piracy attacks both visually and automatically.